Membrane Proteins Are Dramatically Less Conserved ... - Nick Lane

Membrane Proteins Are Dramatically Less Conserved than Water-Soluble Proteins across the Tree of Life Victor Sojo,*,1,2,3 Christophe Dessimoz,2,4,5 Andrew Pomiankowski,1,2 and Nick Lane*,1,2 1

CoMPLEX, University College London, London, United Kingdom Department of Genetics, Evolution and Environment, University College London, London, United Kingdom 3 Systems Biophysics, Faculty of Physics, Ludwig-Maximilian University of Munich, Munich, Germany 4 Department of Ecology and Evolution, University of Lausanne, Lausanne, Switzerland 5 Center for Integrative Genomics, University of Lausanne, Lausanne, Switzerland 2

*Corresponding authors: E-mails: [email protected]; [email protected]. Associate editor: James McInerney

Abstract

Key words: membrane proteins, orthologs, homeostasis, evolution, adaptation.

Introduction

Article

Biological membranes form the boundary between the cell and its surroundings, and their embedded proteins constitute an active link to the environment, with crucial roles in reproduction, bioenergetics, transport, signaling, and catalysis (Mitchell 1957, 1961; Singer and Nicolson 1972; Hedin et al. 2011). Over half of all known drug targets are membrane proteins (Overington et al. 2006). Their study is therefore central to our understanding of the origins and evolution of life, as well as to physiology and medicine. Previous studies have shown that the subcellular localization of a protein is a strong predictor of its evolutionary rate. Extracellular proteins secreted from the cell evolve faster than intracellular proteins in both mammals and yeast, as do the external parts of membrane proteins (Tourasse and Li 2000; Julenius and Pedersen 2006; Liao et al. 2010), but the reasons are unclear. Structural and packing constraints undoubtedly play a role, with the exposure of amino-acid residues to the solvent (Oberai et al. 2009; Franzosa et al. 2013), as well as the sub-cellular localization of the proteins and their portions (Julenius and Pedersen 2006; Liao et al. 2010) being the strongest predictors of evolutionary rate. Membrane proteins also

diverge faster than intracellular water-soluble proteins in parasites, where surface interactions evolve under pressure to avoid detection by the host (Volkman et al. 2002; Plotkin et al. 2004). This pattern may be specific to the “red-queen” dynamics of parasitic interactions, that is, the need for constant adaptation merely to maintain fitness. Taken together, however, these disparate findings suggest that evolution might generally occur faster outside the cell, and hint at the operation of a wider evolutionary mechanism. Here, we test the hypothesis that protein evolution is faster outside the cell as a result of adaptation to changing environments (fig. 1, top). Over evolutionary time, the interior of the cell remains stable compared with the exterior, which is forced to change in response to shifting biogeochemical processes, migration with colonization of new niches, and other biotic interactions. This leads to the faster evolution of secreted water-soluble proteins and outside-facing sections of membrane proteins. The utility of a protein will also depend on the specific environment, potentially leading to greater loss of membrane-bound proteins over time as environments change (fig. 1, middle). We have analyzed large data sets of orthologs to evaluate the conservation of membrane proteins

ß The Author 2016. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons. org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

2874

Open Access

Mol. Biol. Evol. 33(11):2874–2884 doi:10.1093/molbev/msw164 Advance Access publication August 8, 2016

Downloaded from http://mbe.oxfordjournals.org/ by guest on December 8, 2016

Membrane proteins are crucial in transport, signaling, bioenergetics, catalysis, and as drug targets. Here, we show that membrane proteins have dramatically fewer detectable orthologs than water-soluble proteins, less than half in most species analyzed. This sparse distribution could reflect rapid divergence or gene loss. We find that both mechanisms operate. First, membrane proteins evolve faster than water-soluble proteins, particularly in their exterior-facing portions. Second, we demonstrate that predicted ancestral membrane proteins are preferentially lost compared with watersoluble proteins in closely related species of archaea and bacteria. These patterns are consistent across the whole tree of life, and in each of the three domains of archaea, bacteria, and eukaryotes. Our findings point to a fundamental evolutionary principle: membrane proteins evolve faster due to stronger adaptive selection in changing environments, whereas cytosolic proteins are under more stringent purifying selection in the homeostatic interior of the cell. This effect should be strongest in prokaryotes, weaker in unicellular eukaryotes (with intracellular membranes), and weakest in multicellular eukaryotes (with extracellular homeostasis). We demonstrate that this is indeed the case. Similarly, we show that extracellular water-soluble proteins exhibit an even stronger pattern of low homology than membrane proteins. These striking differences in conservation of membrane proteins versus water-soluble proteins have important implications for evolution and medicine.

Membrane Proteins Are Dramatically Less Conserved . doi:10.1093/molbev/msw164

relative to water-soluble proteins across the entire tree of life, to test whether faster evolution outside the cell is driven by adaptation to new environments and functions.

Results Membrane Proteins Are Shared by Fewer Species in All Three Domains of Life

P ¼ 2.881011; r ¼ 0.712). A simple protein–protein BLAST (BLASTp) search (Altschul et al. 1990) against the full nonredundant (nr) NCBI protein database confirms these findings (supplementary fig. S1, Supplementary Material online). This dramatic reduction in the conservation of membrane proteins is widespread across the entire tree of life, but the effect decreases as cellular or organismal complexity increases. Water-soluble proteins have on average 2.7 times more orthologs than membrane proteins in prokaryotes. The factor decreases to 2.4 in unicellular eukaryotes, and to 1.7 in multicellular eukaryotes (fig. 3A; one-way analysis of variance: F(2,61)¼ 21.07; P ¼ 1.1107; x2 ¼ 0.149). Filtering for proteins shared by eukaryotes and at least one of the prokaryotic domains produces the results in figure 3B. While prokaryotes are largely unaltered and the difference between unicellular and multicellular eukaryotes remains, the effect becomes larger overall for eukaryotes. That is, potentially ancestral proteins in eukaryotes (namely with orthologs in either archaea or bacteria) are more likely to be lost if they are membranebound. We performed a logistic regression on the entire precomputed OMA ortholog data set to estimate the probability that any given protein is membrane-bound as the number of clades sharing it increases. We find that the more universal the protein, the less likely it is to be membrane-bound (fig. 4). Since ortholog discovery depends on the successful detection of homologs using tools such as BLAST, the lower homology of membrane proteins we report could have two main causes (fig. 1). First, it is possible that membrane proteins evolve faster and hence their more divergent sequences are picked up less frequently by homology-identification algorithms. Second, some of the absences may be true gene losses, such that the orthologs are not found because they are genuinely no longer there. We show that both mechanisms are at play.

Faster Evolution of Membrane Proteins and Their Outside-Facing Sections

FIG. 1. Two-fold effect of adaptation causes faster evolution of external sections and loss of homology in membrane proteins. Adaptation to new functions and niches causes faster evolution for outside-facing sections (top), potentially contributing to divergence beyond recognition. Other proteins may provide no advantage in the new environment, and could be lost entirely over time (center). For simplicity, the species on the left is assumed to remain functionally identical to the common ancestor (bottom).

To investigate whether the patterns above are due to membrane proteins having a higher divergence rate overall, we calculated Nei’s sequence-diversity measure (p, Nei and Li 1979) for the 228,148 OMA OGs shared by any three or more species. The results confirm previous reports on data sets with more limited phylogenetic ranges (Volkman et al. 2002; Julenius and Pedersen 2006) that membrane proteins diverge more quickly than water-soluble proteins (Welch’s t-test: t ¼ 14.08, df ¼ 14261.09; P ¼ 2.591045; r ¼ 0.12, fig. 5A); this result is consistent across the three domains of life (fig. 5B–D). While the TMHMM algorithm has been shown to infer trans-membrane helical (TMH) regions with very high accuracy (Krogh et al. 2001), discerning the inside- versus outsidefacing aqueous regions of TMH proteins is substantially more challenging. We downloaded the full nonredundant set of sequences and annotations from the trans-membrane protein data bank (PDBTM, pdbtm.enzim.hu) (Tusnady et al. 2004), to assess the evolution of the three main regions of trans-membrane proteins: inside-facing aqueous, membrane2875


To study the evolution of membrane proteins across the tree of life, we downloaded the 883,176 pre-computed ortholog groups (OGs) for all 1,706 species from the three domains of life present in the OMA database (Altenhoff et al. 2015). We separately obtained the full list of 66 species in the EMBL-EBI list of reference proteomes (www.ebi.ac.uk/reference_pro teomes), and extracted the OMA OGs for each protein of each species, where present (supplementary table S1, Supplementary Material online). We classified each protein sequence as either a membrane protein (MP) or a watersoluble protein (WS) using the predictions of the TMHMM algorithm (Krogh et al. 2001). We then determined the number of orthologs found for each protein (i.e., the size of the ortholog cluster, or OG, for each protein) independently for each species. We find that, in all cases of all three domains of life (archaea, Gram-positive and Gram-negative bacteria, as well as unicellular and multicellular eukaryotes), the mean number of orthologs is substantially smaller for MPs than for WSs (fig. 2 and supplementary table S1, Supplementary Material online); that is, membrane proteins are shared by fewer species on average (paired t-test: t ¼ 8.05; df ¼ 63;

MBE

MBE

WS MP

WS MP

100

150

Gram-negative Bacteria

THEKO

BACSU MYCTU STRCO AQUAE BACTN

BRAJA

CHLAA

CHLTR

DEIRA

DICTD

60

Unicellular Eukaryotes

(cont.)

WS MP

10

20

30

40

Gram-negative Bacteria

50

70

METAC SULSO

ECOLI

FUSNN

GEOSL

GLOVI

LEPIN

PSEAE

RHOBA SYNY3

THEMA THEYD

0

0 Mean number of orthologs per protein 10 20 30 40 50 60 70

Gram-positive Bacteria

50 KORCO METJA

0

HALSA

Unicellular Eukaryotes

Multicellular Eukaryotes

(cont.)

WS MP

PHANO

PLAF7

Mean number of orthologs per protein 10 20 30 40 50 60 70

0

200

WS S MP

SCHPO

YARLI

YEAST

ANOGA

ARATH

BOVIN

ASPFU CANAW CRYNJ

GIAIC

LEIMA

MONBE

BRAFL

CAEEL

CANFA

CHICK

CIOIN

DANRE

DICDI

DROME

PANTR

PHYPA

RATNO

SCHMA

SCLS1

TAKRU

USTMA

XENTR

Multicellular Eukaryotes (cont.)

HUMAN

IXOSC

MACMU MONDO MOUSE NEMVE

NEUCR ORNAN

FIG. 2. Membrane proteins have fewer orthologs in all three domains of life. The mean size of OMA Ortholog Groups (OGs) is substantially smaller for membrane proteins in all 64 species in the EMBL-EBI’s list of reference proteomes studied (2 of the 66 species were not found in OMA at the time of this analysis). Five-letter codes are OMA species identifiers; details in supplementary table S1, Supplementary Material online. Dark shade: water-soluble (WS); light shade: membrane proteins (MP). Data represented as the mean number of orthologs that WSs and MPs of each genome have in OMA62SEM (standard error of the mean).

spanning, and outside-facing aqueous. Briefly, this database has annotations, where available, for the sub-cellular localization of each amino acid in all membrane-protein structures deposited in the Protein Data Bank (PDB, www.rcsb.org) (Berman et al. 2000; Rose et al. 2015). We performed a BLASTp search of the sequence of each PDB structure against our subset of the OMA database, aligned the sequences of the best-matching orthologous groups, and sliced the alignments vertically to obtain the inside-facing, membrane-spanning, and outside-facing regions, plus an “aqueous” 2876

assemblage constructed by concatenating the inside and outside portions (see Materials and Methods section for details). We next calculated Nei’s sequence-diversity measure (p) for each section of the protein alignments (fig. 5E). The results confirm that aqueous regions evolve faster than membranespanning regions (paired t-test: t ¼ 8.87; df ¼ 309; P ¼ 5. 951017; r ¼ 0.450). Amongst the aqueous regions, both of which have faster rates than the membrane-spanning regions overall, the outside-facing portions evolve faster than their inside-facing counterparts (fig. 5E; paired t-test: t ¼ 3.76;


0

250

Archaea

Mean number of orthologs per protein 50 100 150 200 250

Mean number of orthologs per protein 0 25 50 75 100

Sojo et al. . doi:10.1093/molbev/msw164


df ¼ 296; P ¼ 2.07104; r ¼ 0.213). These results are confirmed using an additional estimate computed by building trees for the sliced alignment portions and averaging the branch lengths of all nodes within each tree (supplementary fig. S2, Supplementary Material online; see Materials and Methods section for details). As before, aqueous regions are shown to evolve faster than membrane-spanning regions (paired t-test: t ¼ 10.2109; df ¼ 371; P ¼ 1.401016; r ¼ 0. 411), whereas specifically the outside-facing sections once more have faster rates than their corresponding insidefacing sections (paired t-test: t ¼ 4.63; df ¼ 359; P ¼ 5. 22106; r ¼ 0.237). To control for potential errors in the automatic annotations of PDBTM, we repeated our analysis by manually annotating the 3 main regions (inside, outside, and membranespanning) of 12 membrane proteins that are highly shared in OMA, including 1 outer-membrane beta-barrel porin and 11 trans-membrane helical proteins. The closest-matching structural file was found by BLASTp search against the PDB subset on the NCBI server. The subcellular location of each aminoacid residue was then assigned by inspecting the PDB structures against the information in the corresponding primary literature (supplementary table S2, Supplementary Material online). Orthologs were assigned from the corresponding

OMA OG (except in the case of OmpF, whose homologs were obtained from a BLASTp search against the nr database). The homologous sequences were aligned to the PDB sequence, alignments sliced and evolutionary rates estimated using Nei’s p. In 10 of the 12 proteins hand-annotated in this way, evolution occurs faster for outside-facing than for insidefacing aqueous regions (supplementary fig. S3, Supplementary Material online; paired t-test: t ¼ 4.97; df ¼ 11; P ¼ 4.25104; r ¼ 0.832). Using the mean branch lengths of trees as an alternative estimate of evolutionary rates, all 12 proteins show faster rates in the outside-facing regions than in their insidefacing counterparts (supplementary fig. S4, Supplementary Material online; paired t-test: t ¼ 4.71; df ¼ 11; P ¼ 6. 37104; r ¼ 0.818). These findings are again widespread across the tree of life, and apply to multiple types of proteins. We note that these patterns hold true despite the fact that some aqueous proteins are exported from the cell and predictably evolve faster (Julenius and Pedersen 2006), whereas some membrane proteins, especially in eukaryotes, sit on organellar membranes (hence presumably evolve slower).

Extracellular Water-Soluble Proteins Have Fewer Orthologs than Membrane Proteins To estimate the effect of extracellularity, we used the predictions of the SignalP package (Petersen et al. 2011). Briefly, this software detects fragments of amino-acid sequences likely to target proteins to the secretory pathway. SignalP detects these signal peptides in most Gram-positive and Gramnegative bacteria, as well as eukaryotes (note that the software is presently unable to reliably predict signal peptides in archaea). We re-classified all water-soluble OMA OGs as either intracellular (i.e., cytosolic), or extracellular (i.e., containing a signal peptide and not being a membrane protein), based on SignalP predictions. Extracellular proteins are shared on average by notably fewer clades than intracellular (most simply cytosolic) proteins (fig. 6A; mean OG size of intracellular WS proteins 8.60 versus 6.55 for extracellular; Welch’s ttest: t ¼ 27.62, df ¼ 29602.0; P ¼ 7.4010166; r ¼ 0.16). Membrane proteins are intermediate: less widespread than intracellular proteins, but more so than extracellular ones. Similarly, grouping proteins by the proportion of their residues that are exposed to the environment produces a pattern of falling phylogenetic spread as extracellular exposure increases (fig. 6B; linear regression on data binned as described in Materials and Methods; F(1,8)¼25.45; P ¼ 0.00147; r ¼ 0.859). The evolutionary rates, measured by Nei’s p, are faster for extracellular than for intracellular water-soluble proteins (fig. 6C; mean for intracellular WS proteins 0.592 versus 0.610 for extracellular; Welch’s t-test: t¼22.67, df ¼ 17163.09; P ¼ 3.7010112; r ¼ 0.171), whereas membrane proteins show a slightly higher rate.

Membrane Proteins Have Been Lost More Often within Closely Related Species The results in figure 5 suggest that the higher evolutionary rates of membrane proteins could, through divergence beyond recognition, lead to the loss of homology reported 2877


FIG. 3. Water-soluble orthologous groups are substantially larger on average than membrane-protein groups, but the effect decreases as organismal complexity increases. Dividing the average size of watersoluble orthologous groups (OGs) of each species over the corresponding average size of membrane-protein OGs gives an indication of the magnitude of the effect in figure 2 for the different groups of species. (A) The ratio of the mean sizes of water-soluble over membrane protein OGs is > 1 for all species studied (i.e., each WS bar is always larger than its corresponding MP bar in figure 2), but the effect decreases as cellular and organismal complexity increase, from prokaryotes to unicellular eukaryotes, to multicellular eukaryotes. (B) Filtering for orthologous groups composed of both eukaryotes and prokaryotes keeps the relationship between unicellular and multicellular eukaryotes and indeed increases the effect, whereas prokaryotes remain largely unaltered. This suggests that membrane proteins ancestral to eukaryotes (i.e., with ancestors in archaea or bacteria) have been lost more often than their water-soluble counterparts. Bold black lines represent the median, white lines the mean, and boxes and whiskers are standard in R at a 6 1.5*IQR (inter-quartile range) threshold. Numbers below the boxes indicate sample sizes.

MBE

MBE

β0=−1.19 | β1=−0.0069 | t=−14.44 | p=2.86e−47 Any 3+ clades (228,148 OGs)

0.3

%0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

β0=−0.66 | β1=−0.072 | t=−9.72 | p=3.24e−22 Archaea (8,148 OGs)

0.1

0.05 0.1

0.0 200

300

400

%0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

D

10

20

30

40

50

60

70

%0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

β0 =−1.17 | β1 =−0.109 | t=−13.39 | p=8.44e−41 Eukaryotes (36,961 OGs)

0.1

β0 =−1.13 | β1 =−0.0059 | t=−12.23 | p=2.27e−34 Bacteria (181,908 OGs)

0

0.2

100

0.15

C

0.05 0.0

0.0 0

100

200

300

Number of taxa that share the protein

0

5

10

15

20

25

30

Number of taxa that share the protein

FIG. 4. The probability of a protein being membrane-bound falls with wider distribution. (A) A logistic regression shows that the probability that a gene is a membrane protein falls significantly with increasing number of clades sharing it, for OGs shared by any 3 or more separate clades. The pattern remains when considering each of the three domains separately (B–D). The points and vertical stripes correspond to the proportions of MPs amongst genes shared by increasingly large numbers of clades, divided in 10% bins. No proteins retrieved were shared by over 90% of the 489 taxa in (A). In all cases, the final bins have proportion zero, i.e., no highly shared proteins are membrane-bound. Note that the points and bins are provided for reference only: logistic regressions were performed on the individual ortholog clusters (i.e., the probability curves were derived independently, see Materials and Methods section).

earlier (fig. 2 and supplementary fig. S1, Supplementary Material online). To determine whether true gene loss has occurred as well, we repeated the presence–absence analysis (fig. 2) on sets of proteins predicted to be ancestral to closely related species and strains. We selected all prokaryotic clades with 10 or more closely related species in OMA, and assumed that proteins shared by more than half of the members of the clade were ancestral (see Materials and Methods section). We considered that any clades that do not share such ancestral proteins represent true gene losses, on the assumption that in closely related strains and species orthologs are unlikely to have diverged beyond recognition. The results show that the mean numbers of species sharing each of these ancestral OGs are lower for membrane-bound than for water-soluble proteins across 31 of the 35 clades studied (fig. 7; paired t-test: t ¼ 7.31; df ¼ 34; P ¼ 1.81108; r ¼ 0.782). That is, membrane proteins have been lost more often than water-soluble proteins between closely related taxa, confirming that true gene loss can also account in part for the lower homology of membrane proteins reported here.

Discussion We report that membrane proteins have fewer orthologs than water-soluble proteins across the entire tree of life (figs. 2 and 4). In principle this finding could be due to a higher evolutionary rate, which prevents sequence-searching algorithms 2878

such as BLAST from detecting homologs beyond a given threshold, or it could correspond to true gene loss. We show that both mechanisms are at play. First, we demonstrate that evolutionary rates are faster for membrane proteins than for water-soluble proteins across the whole tree of life, and in each of the three domains of archaea, bacteria, and eukaryotes independently (fig. 5A–D). Significantly, the evolutionary rates of membrane proteins are faster in the outside-facing aqueous regions than in their inside-facing counterparts (fig. 5E and supplementary figs. S2–S4, Supplementary Material online). Second, our analysis of closely related species shows that predicted ancestral proteins have been lost more frequently if they were membrane bound (fig. 7). This indicates that the lower homology of membrane proteins is not only due to divergence beyond sequence recognition, but also that true gene loss has occurred. It has been reported that exported water-soluble proteins evolve faster than cytosolic proteins, and indeed faster than the external sections of membrane proteins in mammals (Julenius and Pedersen 2006). Our hypothesis predicts that a similar pattern of low homology and greater gene loss should be observed for excreted water-soluble proteins. We confirm that this is indeed the case (fig. 6A and B), with universality decreasing the further out the cell from cytosolic, to membrane-bound, to extracellular proteins. Membrane-bound and extracellular proteins are in general less central to metabolic networks and functions (Julenius


0.05 0.1

0.15 0.2 0.25

0

Probability that protein is membrane-bound

B

0.2

%0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90

0.15 0.2

0.25

A

0.0

Probability that protein is membrane-bound

Sojo et al. . doi:10.1093/molbev/msw164


and Pedersen 2006), so the patterns we report could be caused by stronger selective constraints operating on cytosolic proteins, and a comparatively relaxed evolution of more peripheral proteins. However, at least for mammals, the faster evolutionary rates of membrane-bound and extracellular proteins do not seem to depend on the essentiality of the proteins themselves (Liao et al. 2010), suggesting that mechanisms other than purifying selection on purportedly less crucial membrane-bound and exported proteins are at play.

Our findings suggest the operation of a more general evolutionary principle: membrane proteins evolve faster because they face stronger adaptive selection in changing environments, whereas cytosolic proteins are under more stringent purifying selection in the homeostatic interior of the cell (fig. 1). The outside-facing sections of membrane-spanning proteins are closely involved in adaptation to new environments and functions, and so are more likely to diverge over time than the cytosolic portions. As emerging species colonize novel environments or specialize in new tasks, the outside-facing sections are subject to stronger positive selection, whereas rate-limiting purifying selection prevails in the membrane-spanning and inside-facing portions (fig. 5E and F; supplementary figs. S2–S4, Supplementary Material online). Novel or changing environments are also likely to reduce the utility of existing membrane proteins, leading to loss over time, and accounting for the absences that we observe in closely related species (fig. 7). Our hypothesis immediately suggests that this effect should be strongest in prokaryotes, weaker in unicellular eukaryotes (where intracellular organelles can provide an additional homeostatic environment for membrane proteins), and weakest in multicellular eukaryotes (where even extracellular proteins face a homeostatic environment provided by tissues and organs). That is indeed the case (fig. 3A), although the difference in size of ortholog groups between membrane proteins and water-soluble proteins remains substantial even in multicellular eukaryotes. Moreover, the difference between water-soluble and membrane-bound proteins is greater for proteins most simply assumed to be ancient to eukaryotes (fig. 3B). This reinforces the suggestion that ancient membrane-bound proteins are more likely to either diverge beyond recognition or be lost entirely than their water-soluble counterparts. This broad evolutionary perspective provides a framework for interpreting a number of earlier findings that have proved difficult to generalize. Previous results show that watersoluble proteins secreted from the cell evolve faster than cytosolic proteins in mammals and yeast, and that the external portions of membrane proteins evolve faster than the internal domains (Julenius and Pedersen 2006). However, given the complexity of mammalian species, a focus on this taxonomic class does not lend itself to generalizations about purifying selection or adaptation to changing extracellular environments. Similarly, the G-protein-coupled receptor superfamily is known to evolve faster in its extracellular portions than in the transmembrane and cytosolic regions, but this has again been interpreted in terms of particular functional and structural constraints (Tourasse and Li 2000; Lee et al. 2003). In Gram-negative bacteria, degradation of xenobiotic toxic substances occurs in the periplasmic space (Kawai 1999; Nagata et al. 1999), making evolutionary pressure stronger on the external regions than in the homeostatic interior. Signal peptides have been shown to evolve rapidly in both prokaryotes and eukaryotes, pointing to positive selection on these secretory membrane-targeting fragments (Li et al. 2009). Finally, parasitic interactions can promote the rapid evolution of membrane proteins, especially the external loops involved directly in antigen interactions (Volkman et al. 2002; Plotkin 2879


FIG. 5. Membrane proteins evolve faster, especially in their external sections. (A–D) Nei’s sequence diversity measure (p) is higher for membrane proteins (MP) than for water-soluble proteins (WS) in the full set of OMA OGs (A) as well as for each of the three domains of life separately (B–D), indicating that evolution occurs faster for MPs. (E) For sections of membrane-protein OMA OGs annotated from the structures in the PDBTM database, Nei’s p shows that aqueous sections evolve faster overall than membrane-spanning sections. Splitting the aqueous sequences into outside- and inside-facing sections confirms that regions exposed to the environment evolve faster than those facing the cytosol. Boxplot ranges as in figure 3 with notches at the 95% confidence-interval around the median. All comparisons of WS to MP in (A–D), as well as inside and outside portions to each other or to membrane-spanning portions in (E) had P